Date: Fri, 25 Nov 2005 11:07:50 +1300
From: Mark Kirkwood <markir@paradise.net.nz>
Subject: Re: Hardware/OS recommendations for large databases (
In-reply-to: <14286.1132849528@sss.pgh.pa.us>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Greg Stark <gsstark@mit.edu>, Luke Lonergan <llonergan@greenplum.com>,
	stange@rentec.com, Dave Cramer <pg@fastcrypt.com>,
	Joshua Marsh <icub3d@gmail.com>, pgsql-performance@postgresql.org
Message-id: <438639B6.5090902@paradise.net.nz>
MIME-version: 1.0
Content-type: text/plain; format=flowed; charset=ISO-8859-1
Content-transfer-encoding: 7bit
User-Agent: Mozilla Thunderbird 1.0.6 (X11/20051106)
References: <BFAAB702.1460D%llonergan@greenplum.com>
	<438583C8.4050809@paradise.net.nz> <87d5kqkpd2.fsf@stark.xeocode.com>
	<14286.1132849528@sss.pgh.pa.us>

Tom Lane wrote:
> Greg Stark <gsstark@mit.edu> writes:
> 
>>Last I heard the reason count(*) was so expensive was because its state
>>variable was a bigint. That means it doesn't fit in a Datum and has to be
>>alloced and stored as a pointer. And because of the Aggregate API that means
>>it has to be allocated and freed for every tuple processed.
> 
> 
> There's a hack in 8.1 to avoid the palloc overhead (courtesy of Neil
> Conway IIRC).
>

It certainly makes quite a difference as I measure it:

doing select(1) from a 181000 page table (completely uncached) on my PIII:

8.0 : 32 s
8.1 : 25 s

Note that the 'fastcount()' function takes 21 s in both cases - so all 
the improvement seems to be from the count overhead reduction.

Cheers

Mark